Enhancing Response Quality by Children in Voice-based Sleep Diaries via AI-based Continuous Feedback

Paper Title

Enhancing Response Quality by Children in Voice-based Sleep Diaries via AI-based Continuous Feedback

Publication Info

• Topic area: Improving children’s response quality in voice-based self-report systems using AI-driven feedback.
• Keywords: Voice-based diaries, continuous feedback, response quality, child-computer interaction, self-reporting, pediatric healthcare, symbolic feedback, numeric feedback, AI in education, conversational agents.

Background and Problem

• Problem / challenge: Children’s responses in self-reporting tools like sleep diaries are often incomplete, vague, or inconsistent, limiting their clinical utility. Sustaining high-quality responses over time is particularly challenging due to children’s developing cognitive and self-regulation skills.
• Significance: High-quality self-reports are critical for accurate clinical assessments, especially in pediatric healthcare contexts like sleep disorder treatment. Addressing response quality can improve data reliability and support better clinical outcomes.
• Motivation and related work: Prior research has shown that feedback can improve task performance and engagement, but its application to children’s voice-based self-reporting remains underexplored. Existing studies highlight the need for child-specific feedback designs that balance clarity and engagement.

Solution

• Proposed approach: An AI-powered voice-based sleep diary with "live", continuous feedback to improve children’s response quality across questions and over multiple days.
• Novelty:
  1. Introduction of real-time, continuous feedback in voice-based self-report systems for children.
  2. Empirical comparison of symbolic (smiley) and numeric feedback on response quality.
  3. Field study demonstrating sustained response quality improvement over multiple days using combined feedback.
• Procedure and key techniques:
  1. Co-design study to explore children’s preferences for feedback types (numeric, symbolic, progress).
  2. Lab experiment comparing the effects of smiley, numeric, and no-feedback conditions on response quality across questions.
  3. Eight-day field study testing a combined feedback system (smiley + numeric) to evaluate sustained response quality over time.

Results

• Concrete findings:
  • Numeric feedback yielded the highest mean response quality (M=3.98), followed by smiley feedback (M=3.75), and no feedback (M=2.81).
  • Continuous feedback improved response quality across questions within a session and sustained or improved quality over multiple days.
  • Feedback conditions showed a positive trajectory in response quality, while no-feedback conditions showed stagnation or decline.
• Advantage over baselines: Both numeric and symbolic feedback significantly outperformed the no-feedback condition in improving response quality. Combined feedback in the field study maintained higher quality over days compared to no feedback.
• Experiments / evaluation:
  • Co-design study with 14 children (ages 7–12) identified preferences for symbolic feedback.
  • Lab study with 36 children (ages 7–12) used a within-subject design to compare feedback types.
  • Field study with 24 children (ages 8–12) employed a within-subjects crossover design over eight days to assess sustained effects.
• Limitations and future work:
  • Limited sample diversity and short study duration; longer-term studies are needed.
  • Lack of direct comparison between combined feedback and individual feedback types.
  • Findings may not generalize to clinical populations or non-Western contexts.

Summary

This paper introduces an AI-powered voice-based sleep diary with "live", continuous feedback to enhance children’s response quality in self-reporting. Through co-design, lab, and field studies, the authors demonstrate that both symbolic (smiley) and numeric feedback improve response quality across questions and sustain it over multiple days. Combined feedback was particularly effective in maintaining engagement and quality over time. The findings highlight the importance of balancing emotional engagement with cognitive clarity and suggest that adaptive, multilayered feedback systems can improve pediatric self-reporting in healthcare and other contexts.